search

fungtion / DSN

Pytorch implementation of Domain Separation Networks
147 stars 38 forks source link

Question about the Sign of SIMSE Loss #9

Open MaxLiu245 opened 4 years ago

MaxLiu245 commented 4 years ago

Hi, I was reading the DSN paper and saw you codes here. Nice codes!

But here I was wondering whether the loss codes in train.py's line 189 & 245 should be loss -= source_simse and loss -= target_simse. To confirm this, I read the original scale-invariant mean squared error paper, especially the eqn. 4 and made this conclusion.

I thought if so, the acc can be higher, and get close to DSN's result~

BX-xb commented 4 years ago

Hi, do you have the code of the original paper? It seems that the github link in the paper cannot be entered

codenie commented 3 years ago

The official implementation of the DSN model is in the archive brach, the link is as follows: https://github.com/tensorflow/models/tree/archive/research/domain_adaptation

Hi, do you have the code of the original paper? It seems that the github link in the paper cannot be entered

codenie commented 3 years ago

In addition to the sign, the log should also be used, according to the equation of SIMSE: image where image Is this right?